A Detailed Review of Computational Methods and Signal Processing Techniques for Automated Detection of REM Sleep Behavior Disorder

Background/Objectives: REM Sleep Behavior Disorder (RBD) is a parasomnia that manifests as the occurrence of abnormal motor movement in the REM sleep and is becoming known as a prominent early indicator of neurodegenerative diseases. Over the last couple of years, due to the increasing need and desire to find efficient and objective means of diagnostics, there has been much research within the area of automated RBD detection. This paper will attempt to give a deep introduction to the contemporary methodologies and technological developments in the automated detection of RBD with the help of physiological signals. Methods: Current literature was also reviewed in a systematic manner, and the emphasis was laid on studies that employ polysomnography data such as electroencephalography (EEG) and electromyography (EMG) to detect RBD. The focus was made on signal processing, feature extraction algorithms, especially spectral and time-frequency features, as well as machine learning and deep learning models application. Unimodal and multimodal approaches were also compared to each other. Results: Available studies reveal that spectral characteristics, particularly the ones obtained using the higher frequencies bands, are significant in discriminating between RBD victims and healthy participants. EMG-based features have high discrimen-native ability because of their capability to obtain an abnormal muscle activity whereas the EEG-based features offer the supplementary information in neural dynamics. Multimodal frameworks that combine EEG and EMG signals have also been shown to enhance accuracy of detection. The use of new models of computation, such as convolutional neural networks and hybrid classifiers, has made additional improvements to the per-performance, yet issues of data variability and generalization persist. Conclusions: Signal processing and smart algorithms have also led to significant development in automated RBD detection. The results of this work are promising but more efforts are needed to work on standardized, reliable and clinically relevant systems. Future studies need to work on increasing the model interpretability, whereas feature selection should be optimized, and need validation on a wide range of datasets to facilitate the implementation in the real world.Overall, automated RBD detection is one of the rapidly developing areas that can help diagnose the disorder earlier, conduct mass screening, and gain a deeper insight into sleep-related disorders.